Sarcasm-Detection-ArSarcasm-Dataset

URL: https://github.com/Moamen-Elsayed/Sarcasm-Detection-ArSarcasm-Dataset

Description: Sarcasm detection from a collection of Arabic tweets.

Dataset

• Name: ArSarcasm-v2
• Size:
  • Training data: 12,548 tweets
  • Testing data: 3,000 tweets
• Fields:
  • tweet: The original tweet text.
  • sarcasm: Boolean indicating whether a tweet is sarcastic or not.
  • sentiment: The sentiment of the tweet (positive, negative, neutral).
  • dialect: The dialect used in the tweet, categorized into the following:
    • msa: Modern Standard Arabic.
    • egypt: Dialect of Egypt and Sudan.
    • levant: Dialect including Palestine, Jordan, Syria, and Lebanon.
    • gulf: Dialect of Gulf countries (Saudi Arabia, UAE, Qatar, Bahrain, Yemen, Oman, Iraq, and Kuwait).
    • magreb: Dialect of North African Arab countries (Algeria, Libya, Tunisia, and Morocco).

Additional Information

How the datasets were created

ArSarcasm-v2 is an extension of the original ArSarcasm dataset, created by combining portions of the DAICT corpus and newly annotated tweets. Each tweet was labeled for sarcasm, sentiment, and dialect. The annotations were performed by native speakers, ensuring high-quality data. The dataset was designed to address challenges in sarcasm detection and sentiment analysis and was released as part of a shared task.

Training methods applied

The dataset utilizes pre-trained embeddings:

• Emoji embeddings: Pre-trained word-emoji embeddings from Emoji Embeddings.
• Word embeddings: Pre-trained word embeddings from Majazak.

Results obtained

Information not available.